The Western Pacific Subtropical High (WPSH) is crucial in shaping summer climate of East Asia. However, our understanding of its seasonal prediction skill is still evolving. Previous studies on seasonal prediction of the WPSH have generally focused on the recent 20–30 years, making it challenging to examine decadal variations in prediction skill. Here, based on a 110-year-long seasonal hindcast (1901–2010) called CSF-20C, we find the prediction skill of summer WPSH has shown a consistent improvement since the 1960s. The enhanced prediction skill can be understood by the increased signal-to-noise ratio of WPSH in the prediction model. The predictable components of WPSH primarily originate from three key oceanic regions: the tropical central-eastern Pacific, the tropical North Atlantic, and the Indo-Pacific. Notably, the Indo-Pacific region contributes the most to WPSH prediction skill, accounting for 53% during high-skill period (1961–2010) and 36% during low-skill period (1901–1950). The enhanced predictable components of the WPSH stem from an increase in observational data and enhanced quasi-biennial variability of ENSO since the 1960s, enabling the model to retain more observed signals in the initial conditions for May. Besides predictable components, unpredictable noises are also derived from CSF-20C members. In the upper troposphere, East-Atlantic-like pattern during high-skill period and a dipole-like pattern in the North Atlantic during low-skill period, both influence the East Asian westerly jet and lead to shifts in the intensity and position of the WPSH downstream. Our study provides greater insight into the mechanisms involved in the seasonal prediction of WPSH.

century-long seasonal prediction,WPSH dynamics,three-ocean interactions,signal-to-noise ratio